P o l i ci e s , P o l i t i c s: C a n E v i d e n c e P l ay a
R o l e i n t h e F i g h t ag a i n s t P ov e r t y ?
Esther Duflo
The Sixth Annual
Richard H. Sabot Lecture
April 2011
The Center for Global Development
The Richard H. Sabot Lecture Series
The Richard H. Sabot Lecture is held annually to honor
the life and work of Richard “Dick” Sabot, a respected
professor, celebrated development economist, successful
internet entrepreneur, and close friend of the Center for
Global Development who died suddenly in July 2005. As
a founding member of CGD’s board of directors, Dick’s
enthusiasm and intellect encouraged our beginnings.
His work as a scholar and as a development practitioner
helped to shape the Center’s vision of independent research and new ideas in
the service of better development policies and practices.
Dick held a PhD in economics from Oxford University; he was Professor
of Economics at Williams College and taught previously at Yale University,
Oxford University, and Columbia University. His contributions to the fields of
economics and international development were numerous, both in academia
and during ten years at the World Bank.
The Sabot Lecture Series hosts each year a scholar-practitioner who has made
significant contributions to international development, combining, as did
Dick, academic work with leadership in the policy community. We are grateful
to the Sabot family and to CGD board member Bruns Grayson for the
support to launch the Richard H. Sabot Lecture Series.
Previous Lectures
2010 Kenneth Rogoff, “Austerity and the IMF.”
2009 Kemal Derviş, “Precautionary Resources and Long-Term Development
Finance.”
2008 Lord Nicholas Stern, “Towards a Global Deal on Climate Change.”
2007 Ngozi Okonjo-Iweala, “Corruption: Myths and Reality in a
Developing Country Context.”
2006 Lawrence H. Summers, “Harnessing the Development Potential of
Emerging Market Reserves.”
Esther Duflo
Esther Duflo is a leading development economist known
for her work applying impact evaluation, randomized
controlled trials, and other field experiments to identify
what works in development interventions.
Duflo is the Abdul Latif Jameel Professor of Poverty
Alleviation and Development Economics at the
Massachusetts Institute of Technology and co-founder
of the Abdul Latif Jameel Poverty Action Lab (J-PAL), a research network
specializing in randomized evaluations of social programs. She was named a
MacArthur Foundation Fellow in 2009 and received the 2010 John Bates Clark
Medal for leading economist under the age of forty.
Her Sabot lecture draws upon her new book with Abhijit Banerjee, a J-PAL
co-founder, titled Poor Economics: A Radical Rethinking of the Way to Fight
Global Poverty. Duflo and Banerjee’s research has led them to identify wholly
new aspects of the behavior of poor people, their needs, and the way that aid
or financial investment can affect their lives.
Duflo completed her undergraduate studies at L’Ecole Normale Supérieure
and holds a PhD in economics from MIT. She is the founding editor of the
American Economic Journal: Applied Economics.
Esther Duflo | 1
Policies, Politics: Can evidence pl ay a
role in the fight against povert y ?1
Most of my work has focused on trying to figure out what policy may work
in helping to address the various problems linked to poverty. To this end, I
have worked on the evaluation of the effectiveness of various policy options,
particularly (but not only) on randomized evaluation. I have also spent a lot of
time trying to understand how people behave, and why they behave the way
they do, at the most micro level. In that, I have tried, as Richard Sabot’s son
just said his father told us, to learn from failures as much as from successes.
In our new book, Poor Economics, Abhijit Banerjee and I trying to dissect
what has been learned, in decades of work on development economics,
about the key problems and issues that affect the economic lives of the poor:
health, education, savings, credit, labor markets, and more. Our view is that
one needs to understand these issues in detail to be able to design effective
policy, and that a key reason why a lot of efforts to help the poor have failed is
because they have been based on a misunderstanding of the problems that they
were trying to get at, and a failure to learn from experience. This may seem
somewhat reminiscent of a management book from the 1950s. But what is
striking is how far away the process of policymaking usually is from this ideal.
For example, countries spend considerable time and effort trying to deliver
food to people even though all the evidence suggests that, except in rather
extraordinary circumstances, the poor are not particularly hungry for more
calories, and not particularly willing to eat much more even when calories
become cheaper or when they become a bit richer. The problem seems more
about finding ways to improve the nutritional quality of what they eat.
The text of this lecture is adapted from chapter 10 of Poor Econonomics: A Radical Rethinking of the
Way to Fight Global Poverty, by Abhijit Banerjee and Esther Duflo (PublicAffairs, 2011).
2 | Sixth Annual Richard H. Sabot Lecture
Another example is education: much effort has been devoted to put kids in
front of a classroom and a teacher, but less to try to understand what, and
how, they should be taught. As a result, most kids are in school, but only
about half can read anything. This is a huge waste of time, resources, and—
most of all—talent. And yet, in principle, it can be fixed.
Many people are suspicious of these efforts. Bill Easterly once wrote of
randomized control trials: “RCTs are infeasible for many of the big questions
in development, like the economy-wide effects of good institutions or good
macroeconomic policies.” Then, he concluded that “embracing RCTs has
led development researchers to lower their ambitions.”1 This statement was
a good reflection of what we call the “institutionalist view,” a view that is
popular among many scholars and policymakers today. The real problem of
development, in this view, is not one of figuring out good policies: it is to sort
out the political process. If the politics are right, good policies will eventually
emerge. And conversely, without good politics, it is impossible to design or
implement good policies, at least not at any scale.
The current events in Tunisia, Egypt, Libya, and Syria certainly lead one to
think about this: isn’t it the case that the revolutions, whatever their fate, are
the events that are going to have first-order effects on those countries, rather
than any effort we could have been engaged in to, say, improve how the labor
markets works there? Then there is Côte d’Ivoire: how much good would it do
today to know the best way to convince parents to immunize their children?
Isn’t the problem more that we need a functioning government there?
This view of institutions can easily lead to pessimism: as our colleagues Daron
Acemoglu and Jim Robinson have argued, institutions have a tendency to
1. William Easterly, “Development Experiments: Ethical? Feasible? Useful?” blog posted to Aid
Watch July 16, 2009, http://aidwatchers.com/2009/07/development-experiments-ethical-feasibleuseful/.
Esther Duflo | 3
persist.2 Current leaders that have inherited poor institutions will have the
desire (and the ability) to keep them. We don’t really know how institutions
can change: episodes of wholesale institutional changes are few and far
between, and we don’t know how to predict them, let alone what would foster
them.
I do not want to minimize the role of politics. But what I would like to
argue today (based on the last chapter of our book) is that the primacy of
politics over policy that is at the heart of the institutionalist view is perhaps
overemphasized. What we term the political-economy view is that without
good politics, there won’t be good policies and, conversely, that good policies
will follow from politics. I want to argue that neither is true. It might even
be possible to reverse the argument: perhaps good politics will follow good
policies.
I’ll proceed in three steps. First, I will try to argue that perhaps institutions
are not qualitatively different from any other form of policy: like everywhere
else, details matter. Within a broad frame, there is considerable scope for
improvement in how institutions work in practice. Second, I will argue that,
even given a set of institutions, there is usually considerable scope to improve
policy. Third, good politics can sometimes follow from good policy, easily
setting off a virtuous circle.
INS TIT U TIONS versus institutions
The political-economy view starts from a broad definition of institutions as
rules of engagement: democracy or autocracy, property rights, and so on. And
indeed, broad institutional changes are few and far between: we rarely see
successful major transition to democracy; we rarely see a complete cleanup
of corrupt countries. However, these broad INSTITUTIONS, in capital letters,
2. A great overview of their thinking is available in their forthcoming book Why Nations Fail
(Crown publishers).
4 | Sixth Annual Richard H. Sabot Lecture
define only very broad frameworks. Within them, institutions (lowercase)
are actually incarnated on the ground with many very specific rules. There is
considerable scope for improvement and changes, both within generally “bad”
institutions and, perhaps more importantly, within generally good ones.
Improvements within “bad” INS TIT U TIONS
Let’s first look at the changes within generally bad broad INSTITUTIONS. My
first example is from a paper by Monica Martinez-Bravo, Gerard Padro-iMiguel, and Nancy Qian, who study the role of democratic reform in China.3
Although wholesale democratic reforms are few and far between, there are
many instances where democracy has been introduced, to a limited extent and
at the local level, within an authoritarian regime. Electoral reforms have even
taken place in otherwise authoritarian states such as Indonesia under Suharto,
Brazil during the military dictatorship, and Mexico under the Institutional
Revolutionary Party (PRI). More recently, local elections have been introduced
in Vietnam in 1998, Saudi Arabia in 2005, and Yemen in 2001. These reforms
are very imperfect: the elections are often rigged, and the elected officials have
very limited powers.
Yet, Qian and others show that even very imperfect local elections can make
a substantial difference in how local governments are run. In the early 1980s,
village-level elections were progressively introduced in rural China. Early on,
the Communist Party still decided who was allowed to run. The Communist
Party branch continued to operate in the village, with its appointed secretary.
Ballots were not always anonymous, and the ballot boxes were reportedly often
stuffed.
Despite these shortcomings, the authors found that the reforms had a
surprisingly large effect, suggesting greater accountability to villagers. After
3. Monica Martinez-Bravo et al., “Accountability in an Authoritarian Regime: The Impact of
Local Electoral Reforms in Rural China,” manuscript, Yale University (2010).
Esther Duflo | 5
a village starts holding elections, the village chiefs are more likely to relax
unpopular central policies, such as the one-child policy. The reallocation of
farmland, which happens from time to time in Chinese villages, is more likely
to benefit “middle-class” farmers. Public expenditures are more likely to reflect
villagers’ needs.
Another example is corruption. Is it possible to fight corruption with “small”
reforms or do you need to fix the entire system? Take Indonesia for example.
It is one of the most corrupt countries in the world. In 2010, it ranked 110th
out of 178 countries in Transparency International’s Corruption Perceptions
Index.4 In a randomized field experiment there, Benjamin Olken of Harvard
University found that theft was rife on locally managed projects: 20 percent of
the materials and 27 percent of the wage bill was stolen by local contractors or
the village teams.5 This means that the road are fragile, likely to break down at
the first big rains.
Can this local corruption be uprooted without affecting the deeper institutions
that make Indonesia such a corrupt country? Ben Olken decided to try this
out by running an experiment in about 600 villages. In some of them, the
local teams were warned that their account would be audited. Importantly, it
was not a shock team from the World Bank that was going to audit them, but
the same government auditors, likely to be corrupt themselves. Who was going
to monitors the monitors? Was this not doomed to failure? And yet, in villages
were the threat of audit was present, theft declined by one third.
A third example of changes at the margin is given by the success of simple
programs to address so-called ethnic voting. One reason why democracies fail
in Africa, it is often argued, is that voting is often based on ethnic loyalties,
which means that the candidate from the largest ethnic group often wins,
4. See www.transparency.org/policy_research/surveys_indices/cpi/2010/results.
5. Benjamin A. Olken, “Monitoring Corruption: Evidence from a Field Experiment in Indonesia,”
Journal of Political Economy 115(2):200–249.
6 | Sixth Annual Richard H. Sabot Lecture
whatever his intrinsic merit. The traditional view in political science and
political economy is that this is the product of deep social forces, or a social
equilibrium that is hard to change. Systemic changes would be needed to fix
this. Is that really the case?
In Benin, Leonard Wantchekon of NYU ran a pair of experiments that shed
very interesting light on this phenomenon.6 I’ll mention one experiment now
and the other much later in the talk. The first experiment clearly establishes
that ethnic voting is a problem in Benin. The experiment took place in
relatively stable constituencies for some of the major candidates. In all of
them, pairs of villages were randomized, and different discourses were given
during the village meetings.
The clientelist message read as follows:
We are the representatives of the candidate Saka Lafia, who is running
for president in the March 3, 2001, election. As you know, Saka is the
only Bariba candidate, actually the first since 1960. Saka is running
because the northeast region, Borgou-Alibori, is very underdeveloped:
low literacy rates, poor rural infrastructure and health care, etc. . . . If
elected, he will help promote the interests of the Borgou-Alibori region
by building new schools, hospitals, and roads and more importantly,
hiring more Bariba people in the public administration.
And the public-policy message read:
We are the representatives of Saka Lafia, our party the UDS stands
for democracy and national solidarity. Saka is running against the
opposition candidate in the North. If elected, he will engage in a
nationwide reform of the education and health care systems with
6 Leonard Wantchelon, “Clientelism and Voting Behavior: Evidence from a Field Experiment in
Benin,”World Politics 55(3):399–422.
Esther Duflo | 7
emphasis on building new schools, new hospitals, and vaccination
campaigns. In conjunction with other opposition leaders, we will fight
corruption and promote peace between all ethnic groups and all the
regions of Benin.
The clientelist message carried almost 80 percent of the vote; the public-policy
message, just under 60 percent.
Keep this experiment in mind, because we will return to another experiment
in Benin a little bit later in the talk. But in the meantime, let’s go to Uttar
Pradesh to investigate whether ethnic voting is particularly entrenched.
Would people be willing to stop voting on caste issues if they were asked not
to? This seems a bit preposterous, but an experiment by Abhijit Banerjee,
Rohini Pande, and Don and Jennifer Green suggests that it may not be that
difficult to affect ethnic voting in Uttar Pradesh, where caste-based voting is a
defining feature of the political system.
The researchers worked with an NGO that went to some villages and simply
asked people to vote not by caste, but on the issues. And, remarkably, people
were convinced. The simple message, which had no specific information on
any of the candidates, was conveyed through puppet shows, discussions, and
children plays; it reduced the probability that voters would choose a candidate
from their own caste from 25 percent to 18 percent.7
7. Abhijit Banerjee at al., “Can Voters Be Primed to Choose Better Legislators? Experimental
Evidence from Rural India,” working paper (2009), http://weber.ucsd.edu/~jlbroz/PElunch/
Pahnde_05_10_pande.pdf.
8 | Sixth Annual Richard H. Sabot Lecture
Scope of improvement within “good”
institutions
Finally, a huge margin for action arises from the fact that even “good” broad
institutions do not function all that well in practice. There is tremendous scope
for improving them. Brazil provides a dramatic example, as Thomas Fujiwara
describes in a remarkable paper.8 Brazil’s democracy is particularly vibrant,
with many candidates contesting elections. In the elections Fujiwara describes,
this had the paradoxical result of making voting a very difficult operation.
Voters, many of whom had problems with reading and writing, had to choose
a candidate and write his or her name or candidate number on a ballot. As a
result, many of the votes were discounted because they were invalid.
At some point, Brazil introduced electronic voting, not particularly to address
this problem, but because it made it easier to count the votes. Nonetheless, it
became much easier for people to choose their candidates. The system was first
introduced in large constituencies and then rolled out everywhere. Thomas
Fujiwara uses that roll out to investigate the effect on the elections.
The reform had a dramatic effect on the number of invalid votes, which
essentially went to zero. Further, the “new” votes were mainly coming from
the poor, and going to candidates representing the interest of the poor. And
policies became more pro-poor in this area.
Brazil, a country which is now doing a lot for the poor, was unintentionally
disenfranchising about 10 percent of its population simply through “technical”
errors: this gives us the measure of how much can be improved with politics,
even in “good” institutions. We have many more examples in the book, but
I want to shift to the second step in my demonstration: even we take the
institutions as given, there is considerable scope for improving policies.
8. Thomas Fujiwara, “Voting Technology, Political Responsiveness, and Infant Health: Evidence
from Brazil,” mimeo, University of British Columbia (2010).
Esther Duflo | 9
There is Room for New policies w ithin
e xisting institutions
I will spend less time on this now, although this is of course a recurrent topic
in the book. Even if we go down to the details, we find that institutions are far
from fully determining the political outcomes. First, good policies sometimes
happen in bad political environment: it is in Suharto’s Indonesia that 62,000
primary schools were built with oil money, leading to increases in wages and
reductions in fertility and infant mortality for the children who attended these
new schools.9
More importantly, plenty of bad policies happen in good policy environments,
not because of any great conspiracy to keep the poor poor, but simply
because, in practice, policies don’t just magically emerge out of the democratic
process. Policymaking does not happen in the void. Policies are written,
and implemented, by men and women who are sometimes without good
intentions but usually perfectly willing to do the best they can. But like
anything else, policymaking is difficult: it requires a very good understanding
of a problem. We usually get it wrong. Policymakers, like anyone else, are
often subject to the temptation of “lazy thinking.” And unlike in business,
there is no market test to know the impact of the policy in advance.
This is one important reason why government programs (and similar programs
run by NGOs and international organizations) often do not work. The
problem is inherently difficult and the details need a lot of attention. Failures
are often not due to sabotage by a specific group, as a lot political economists
would have it, but because the whole system was badly conceived to start with,
and no one has taken the trouble to fix it.
9. Esther Duflo, “Schooling and Labor Market Consequences of School Construction in
Indonesia: Evidence from an Unusual Policy Experiment,” American Economic Review 91(4):
795–813.
10 | Sixth Annual Richard H. Sabot Lecture
We call this the three-I problem: ideology, ignorance, inertia. Schemes
conceived of in ideology (some cartoon-character view of what the poor
need or want) and in ignorance of the realities in the field persist out of
sheer inertia; once they exist, they are difficult to get rid of. We encounter
the three-I problem routinely in our work. We were especially struck by it
when we decided to start a project on village education committees in Uttar
Pradesh. Every village in India now must have a village education committee:
beneficiary participation is what the World Bank recommends to fix broken
public services, and India has embraced it with enthusiasm.
We went to hundreds of villages in Uttar Pradesh and asked about people’s
views of and experiences with the village education committees. Only 8
percent of villagers knew that their village education committee existed;
only 2 percent knew what they were supposed to do; and 25 percent of the
committee members did not know they were members.10 The committee
was totally disengaged with the school. Efforts by Pratham, a large NGO, to
reinvigorate them were completely useless. And yet, at the same time, Pratham
managed to persuade scores of volunteers to teach children for free outside of
school hours.
We once asked the education secretary in Lucknow about the scheme. We
were told that the committee must have “the parents of the best child in the
school, the worst child in the school, and the parent of the handicapped
child.” This was the same secretary who noted that exams had been abolished
in primary school. She had no idea of how the villages or the teachers were
supposed to implement this rule.
This program suffered from the classic three-I problem. Inspired by an
ideology—people’s power is good—and designed in ignorance of what people
want and how the village works, it was, by the time we were studying it,
10. Abhijit Banerjee et al., “Pitfalls of Participatory Programs: Evidence from a Randomized
Evaluation in Education in India,” American Economic Journal: Economic Policy 2(1): 1–20.
Esther Duflo | 11
entirely sustained by inertia. No one had paid any attention to it for many
years, except for some bureaucrat somewhere who was making sure that all the
boxes had been checked.
Improving policies could lead to improving
politics
Good politics may or may not be necessary for good policies; it is certainly
not sufficient. Large-scale waste and policy failure often happen not because of
any deep structural problem, but because of lazy thinking at the stage of policy
design. There is thus a tremendous amount of slack to improve policies—by
spending more time on designing them and by constantly evaluating them.
How to do that is what we devote most of the book to.
But to conclude, I would like to propose a reversal to the primacy of politics
over policy. I often have the feeling that, if politics is so depressing in
developing countries, it is because voters feel that they have to choose between
bad and worst: this may explain why, by default, they vote for someone from
their caste. No one is in a position to credibly offer them an alternative.
In Leonard Wantchekon’s Benin experiment, for example, perhaps voters
simply did not believe the rather bland “public interest” message. Perhaps they
would have if the message had been backed by some real proposals and did not
ring a little bit empty.
In fact, we don’t need to say “perhaps” because Wantchekon ran a second
experiment, during the next presidential election, that demonstrates exactly
that.11
11. Leonard Wantchekon, “Can Informed Public Deliberation Overcome Clientelism?
Experimental Evidence from Benin,” manuscript, New York University (2009).
12 | Sixth Annual Richard H. Sabot Lecture
With his think tank and friends in Benin, he organized a large policy
conference—one of those “expert” conferences that would probably make
Bill Easterly recoil (it even had international experts, although they were
mainly from neighboring countries). The conference was called “Election
2006: What Policy Alternative?” There were four panels on education, public
health, governance, and urban planning. Four experts (two from Benin and
two from neighboring Niger and Nigeria) provided a white paper with policy
recommendations.
After the conference, parties volunteered to use the recommendations as a
policy platform as part of an experiment. Again focusing on relatively safe
seats, the parties grouped villages in pairs. In some villages, they conducted the
usual festive meetings, complete with balloons, T-shirts, and clientelist appeal.
In others, they ran “town hall meetings” in which the proposals were discussed
in a serious manner.
And this time, the results were reversed: it is the reasoned policy platform that
got more votes, not the clientelist appeal. Good policies can improve politics
because voters, even the poorest, have the desire and the ability to engage with
policies.
Conclusion
The message in conclusion is one of hope, but also of responsibility. The
problems that face the poor are incredibly difficult to fully grasp and
understand. Mostly, we get it wrong, and as a result the solutions we come up
with are disappointing, and progress is slow.
But we cannot hide behind fatalism as an excuse to do nothing: neither
bad politics nor bad policies are fatalities of poor countries. Real politics
can be improved, step by step. Policies can be improved even within given
institutional environments. And improving policies may well be the first step
to improving politics.
Esther Duflo | 13
They both can be improved, here and now, perhaps not everywhere in a
revolutionary ardor like the one we are seeing in the Middle East, but step
by step, in a quiet revolution. Where the responsibility lies is with everyone.
Everyone has a role to play in this process. It is not someone else’s job to figure
it out, or let it happen. It is our collective duty, here in the West as well as in
developing countries.
The Center for Global Development
The Center for Global Development works to reduce global poverty and
inequality through rigorous research and active engagement with the policy
community to make the world a more prosperous, just, and safe place for us
all. The policies and practices of the rich and the powerful—in rich nations,
as well as in the emerging powers, international institutions, and global
corporations—have significant impacts on the world’s poor people. We aim to
improve these policies and practices through research and policy engagement
to expand opportunities, reduce inequalities, and improve lives everywhere.
By pairing research with action, CGD goes beyond contributing to knowledge
about development. We conceive of and advocate for practical policy
innovations in areas such as trade, aid, health, education, climate change, labor
mobility, private investment, access to finance, and global governance to foster
shared prosperity in an increasingly interdependent world.
Board of Directors
Edward Scott*
Chair
Nancy Birdsall*
President
Tomothy D. Adams*
C. Fred Bergsten
Henrietta H. Fore
David Gergen
Thomas R. Gibian*
C. Boyden Gray
James Harmon
Bradley Horwitz
Enrique V. Iglesias
Kassahun Kebede
Mark Malloch-Brown
Edward E. McNally*
Robert Mosbacher, Jr.
Ngozi Okonjo-Iweala
Lant Pritchett (ex officio)
Sheryl Sandberg
S. Jacob Scherr
Smita Singh
Patty Stonesifer
Belinda Stronach
Lawrence H. Summers
Toni G. Verstandig
Adam Waldman*
Honorary Members
John L. Hennessy
Sir Colin Lucas
Amartya K. Sen
Joseph E. Stiglitz
* Member of the Executive
Committee
Advisory Group
Masood Ahmed
Jere Behrman
Thomas Carothers
Kemal Derviş
Shanta Devarajan
Esther Duflo
William Easterly
Kristin Forbes
Carol Graham
Simon Johnson
Anne Krueger
Gina Lambright
Nancy Lee
David Lindauer
Mark Medish
Edward Miguel
Jonathan Morduch
Deepa Narayan
Jane Nelson
Rohini Pande
Kenneth Prewitt
Lant Pritchet (Chair)
Raymond Robertson
Dani Rodrik
David Rothkopf
Rebecca Thornton
John Williamson
Eric Werker
Ngaire Woods
Ernesto Zedillo
Independent research and
practical ideas for global prosperity
www.cgdev.org
1800 Massachusetts Ave, NW
Third Floor
Washington, DC 20036